HN user
More broadly, but not unconditionally. They can apply reasonable (definition of that can be another point of contention) restrictions via their app review process as long as the restrictions are applied consistently to all applicants.
There are several reasons for a platform to not want to offer 3p interop, from security to privacy to competition, but the DMA is targeted EU regulation to mandate interop specifically for the big tech cos’ platforms and this is one of the first major tests of the law.
Meta’s request for interoperability is regarding their hardware, not their flagship apps, if you take it at face value.
They want their glasses and headsets to integrate as tightly as an Apple Watch or Vision Pro to show messages and other notifications, connect to WiFi, and share files with an iPhone, but Apple uses private APIs for their own devices. Meta says that is anticompetitive and the APIs should be public.
Apple lays out their rebuttal in detail here and it’s clear Meta asked for everything in the hopes that Apple will settle for some of these things: https://developer.apple.com/support/downloads/DMA-Interopera...
If AI can help an advertiser get the outcome they want in 1000 ad impressions instead of 5000 ad impressions (with more precise targeting or more appealing creative), Meta can 5x the price of those 1000 ad impressions, keep the advertiser happy with the same ROI they had before, and sell 4000 additional impressions to someone else as incremental profit.
Or they could 4.9x the price to show the current advertiser a better ROI and encourage higher spend from them, while still having the 4000 extra impressions to sell.
The financial companies got higher performing ads delivered to the right people, as well as the ability to measure the actions taken from the ads. In exchange, they gave away your data for free.
Every company running ads on the internet is making this value exchange, but these financial companies screwed up by pouring protected data into Google's and Meta's dumb pipes.
They don't. There is no semantic value to this data for them. To Google and Meta, it's an arbitrary data point. To tax companies, it is data that can utilized to either target the delivery or measure the performance of their own ads.
That said, there doesn't need to be semantic value for this to feed into an ML model that ends up associating tax data attributes with other attributes.
The EU issued fines to Google and Meta for billions of dollars a few weeks before they decided this data transfer would be allowed. Wonder if they'll actually be required to pay the full fines now.
EU customer data must be stored and served within the EU, and that guarantee can't be made with a federated service.
Businesses choose to send this data to the ad platforms for their own benefits - better targeting, measurement, and ML optimization of their ad campaigns.
The businesses are legally accountable for the data they're sending and complying with privacy laws, but to most platforms it's a dumb pipe for whatever data the business chooses to send.
Meta offers their own...it's not hidden.
https://chrome.google.com/webstore/detail/facebook-pixel-hel...
But this doesn't cover server-side data transfer. https://developers.facebook.com/docs/marketing-api/conversio...
"Decimated" is a bit of an overstatement for a business on track for over $!00b in revenue and $40b of profit this year. The growth rate took a huge hit, but it's not like they're on the verge of collapse.
These screenshots show the difference: https://twitter.com/rohankatyal29/status/1440726365402120192
Apple enables system wide tracking of all app usage, installs, and in-app purchases for themselves with a prompt to "Turn On Personalized Ads" while third-parties must show a prompt that says "Ask App Not To Track". For third-party ads to operate effectively, you need to reject the "Ask App Not To Track" prompt in both the app that displayed the ad and the app that was being advertised.
The antitrust probe is not strictly regarding the privacy changes themselves, but that Apple did not subject themselves to the same new policy and opt-in UI. They've preferenced their own services and built a multi-billion dollar ad network using data that's dark pattern opted-in for themselves and dark pattern opted-out for third-parties.
In isolation, this may seem like a standard competitive advantage for a platform, but Apple has drawn attention to their own power by wiping out nearly a trillion dollars of public market cap from other tech and ecommerce companies with the change.
Not coincidentally, Apple quietly announced that they are relaxing some of their privacy restrictions during WWDC last week.
The ruling is focused on Meta, but the order establishes a precedent that applies to ALL American companies. I'm sure Facebook, Google, and Twitter can eventually re-architect their systems and data centers in some way to comply with this, but this is going to be an expensive and technically complex piece of regulation for your standard American company to comply with if they want to operate in the EU.
IAB Europe is the entity being fined, not their participating partners. The linked PDF says their fine is 250k euros, which is "proportionate to the infringment" and less than the maximum.
The data being collected is a random identifier and a consent opt-in for tracking/cookies. These businesses use this framework as their source-of-truth declaring whether an internet user consented to cookies or tracking, and their ad systems behave according to the user's preferences.
So now they're being asked to delete their records of who opted in or out, because that data was illegitimately acquired.
[edit] This could also have implications regarding data collected through other systems based on the assumption that an opt-in was valid.
In response to: how did Apple do this?
Mobile app advertising on iOS has long relied on Apple's IDFA - Identifier for Advertisers - a random, device-unique, resettable identifier that was available to all apps.
An app like Candy Crush would tell ad networks like FB that "IDFA 123456 just installed the app" or "IDFA 654321 just made a $5 in-app purchase." FB could then recognize that IDFA as a user who installed the FB app, and they could use that data to target, optimize via ML, and measure the audience and results of Candy Crush's marketing campaign.
Apple's recent iOS update made IDFA an app-level opt-in, so in order for this existing advertising mechanism to work, you needed to opt-in to tracking on FB (so they could associate your IDFA to a FB user) AND to opt-in to tracking on the other apps (so they could send ad networks data with your IDFA). This double opt-in obviously has a super low rate.
FB wasn't impacted nearly as much as other advertising businesses because they already have identifiers for their users (email, phone number, etc), have a ton of first-party interest/behavioral data, and they mostly run ads on their own properties. Apps who used third-party ad networks to sell and place ads in their apps had their business completely destroyed.
I think this is all fair in terms of user choice and to give Apple control of their own platform, but it opens Apple up for a lot of regulatory scrutiny given that they were able to flip a switch that annihilated a $50b mobile advertising industry, while making their own equivalent data collection opt-out (vs their competitors' opt-in). Even then, Apple frames their data collection option as "Personalize your experience" while forcing other apps to use the phrase "Allow this app to track you." Apple's own ads product has tripled its market share since this launched. [0]
[0] https://www.ft.com/content/074b881f-a931-4986-888e-2ac53e286...
And non-Amazon ecommerce sales (*according to one major third-party warehouse company) went down 27%.
There are many immoral players in the industry, but there's nothing inherently immoral about advertising.
I help people discover products and services that they love.
I help thousands of business owners find more customers, so they can support their families and create jobs for others.
There's a lot of bad behavior from marketers and sketchy business owners, but I know the work that I'm personally doing has a tangible, positive impact on the lives of thousands of people.
The data is sent from the advertiser to the ad platform, not from the publisher to the ad platform. The advertiser is incentivized to send accurate data for both performance optimization and for campaign measurement purposes.
Ad fraud is a real problem in the ecosystem, but the server-side APIs are actually more secure. You have a private signed backend endpoint rather than public JS that can be injected anywhere and fed fake data by a malicious party.
I couldn't tell you a specific site, but here's an explanation of both Google's and Facebook's functionality.
FB: https://developers.facebook.com/docs/marketing-api/conversio...
Google: https://developers.google.com/google-ads/api/docs/conversion...
This is also why companies like Tealium and Segment are currently valued at billions of dollars. They provide a single middleware integration point to funnel customer data to the dozens of marketing companies that are now leveraging server-side APIs instead of browser pixels.
This is cool, but not super useful in 2021.
Google and Facebook have already moved their tracking technologies beyond frontend network calls due to a rise in browser-level blocking (browser security policies, international regulations, AdBlock, PiHole, etc). The next generation of tracking tech relies on the backend transfer of data between a website and the ad platform, which is invisible to your own network.
It also shifts the story from "Google and Facebook nonconsensually tracking your every digital move through websites and applications" as described in the article into "websites and applications actively transmitting customer data to Google and Facebook." The websites and applications are no longer passive partners, and they assume the responsibility of managing user consent.
Last quarter, they made $10.12 per user[1, slide 4], however that's not evenly distributed. They made $53.01 per user in US and Canada. But even that isn't evenly distributed. The users who are likely to pay for the subscription are also the most valuable to advertisers because they have disposable income. Some Instagram users see ads for luxury cars, real estate, and vacations, while other users see ads for dollar stores and dropshipped junk. FB earns much more in their auction from that first set of users, but those will be the first users willing to pay to opt out of ads.
Dynamic pricing would probably offend both audiences - "what do you mean the algorithm thinks I'm worthless?" "what do you mean that I'm worth thousands of dollars to you?"
[1] https://s21.q4cdn.com/399680738/files/doc_financials/2021/q2...
Facebook has two types of stock. Class A and Class B. Class B stock gets 10 votes per share, while Class A stock gets 1 vote.
Zuckerberg and other very early people had Class B stock so they can continue to have more than 50% of the vote even when they own less than 50% of the stock.
This was a really popular stock structure in Silicon Valley for a while because it cements founders in control - Evan Siegel at Snap has the same deal, and Adam Neumann at WeWork was about to do it. But now there's a lot of push back against it, and some indexes like the S&P 500 refuse to list new stocks with this dual class structure.
The marketing world is currently a game of musical chairs with buyers and sellers switching sides or people making lateral jumps to different clients or agencies.
The cost of switching is lower with remote work - fewer social bonds with coworkers, fewer fringe benefits to consider (travel, fancy office, expense/entertainment accounts), no change in commute.
People are willing to try the same job in a new company for the shot at an easier to climb corporate ladder because the only cost to switching is swapping one company-owned laptop and phone for another.
The laws are around consent and the ability to revoke it, which can vary by the continent/country/state/city you're in as well as some other characteristics that may not be known at the time that the client-side code would execute.
Server-side allows businesses to defer the data transfer until it's known whether specific consent is granted or revoked. It also allows you to more easily keep a record of the data shared with other parties in the event that a user withdraws prior consent or invokes a right to be forgotten. You then have the ability to tell your partners to also delete those data points.
To play devil's advocate, what if you trusted the ad platform (ha! bear with me. it's a hypothetical) to use that data point to block alcohol advertisers from showing ads to a recovering alcoholic.
In a world with personalized advertising, you can avoid those ads forever. In mediums without personalized advertising, you can't watch a sporting event without seeing a beer ad.
Then again, personalized exclusion opens the door to many types of discrimination - like not showing job or housing ads to people with certain traits.
All that to say, the tech can be used for both good and evil. It needs to be managed by regulation, and not by technical feasibility.
No, but it will decrease its effectiveness to some extent. One of the reasons this tech is gaining adoption is that it removes ad tech's dependence on 3P cookies.
For those who are referencing tracking pixels and say this isn't anything new, it's important to understand that this is the next evolution of that technology as a result of government regulations around privacy and browser/OS ecosystem changes from Google, Apple, and Mozilla. This data transfer won't occur at the browser layer for much longer - it won't be as technically feasible without cookies and it may not be legal given consent/opt-out requirements. Instead the data will pass directly from the advertiser's server to the ad platform's server. Of course this tech has existed for a long time - it's a basic API call - but it hasn't been widely adopted in the ad tech industry, while millions of websites are using tracking pixels.
It is not a 1:1 replacement for tracking pixels and lacks some of those creepy features (you're unlikely to get tagged if you simply browse a website without giving up any personal info), but it offers new ones as well (the ability to send arbitrary data to an ad platform).
It doesn't.
For the ad platforms, this lets them optimize their ads for better performance when they know which user profiles converted.
For advertisers, it's used for directional guidance on the platform, e.g. ad campaign A converts at 3x the rate of ad campaign B.
The famous quote in the industry is "Half the money I spend on advertising is wasted; the trouble is I don't know which half." This method gives you a better idea of which money is wasted, at least compared to something like a TV or a print ad.
It gets more complex when you're advertising on multiple channels. For instance, if you see an ad on FB, Google the product, then buy it, both will take credit for a conversion even though your business only had one sale. There are more scientific methods for modeling advertising results from multiple channels [1], that leverage control groups (say you run a Google ad in New York and a TV ad in California and monitor which market sees a bigger spike in sales).